Stochastic Systems Group  

Energy Scaling Laws for Distributed Inference in Random Fusion Networks
Animashree Anandkumar
Cornell University
Dependency graph is an inherent property of the nodes in a network which can be inferred through the data they produce. However, communication between the nodes consumes energy, and communicating all the raw data for inference is not scalable. In this case, the average energy consumption at a node becomes unbounded as the network grows. We propose a scalable communication scheme through distributed computation of a sufficient statistic for inference. Our scheme has strictly bounded average energy consumption even as the network grows, for a class of dependency models.
Problems with this site should be emailed to jonesb@mit.edu